Cross-Validation of Two Accelerometers for Assessment of Physical Activity and Sedentary Time in Preschool Children

Purpose: The purpose of this study was to cross-validate previously developed Actiwatch (AW; Ekblom et al. 2012) and AcitGraph (AG; Sirard et al. 2005; AG-P, Pate et al. 2006) cut-point equations to categorize free-living physical activity (PA) of preschoolers using direct observation (DO) as the criterion measure. A secondary aim was to compare output from the AW and the AG from previously developed equations. Methods: Participants' (n = 33; age = 4.4 +/- 0.8 yrs; females, n=12) PA was directly observed for three 10-min periods during the preschool-day while wearing the AW (nondominant wrist) and AG (waist). Device specific cut points were used to reduce the AW-E (Ekblom et al. 2012) and AG (AG-S, Sirard et al. 2005; AG-P, Pate et al. 2006) data into intensity categories. Spearman correlations (r(sp)) and agreement statistics were used to assess associations between the DO intensity categories and device data. Mixed model regression was used to identify differences in times spent in activity intensity categories. Results: There was a significant correlation between AW and AG output across all data (r(sp) = 0.41, p <.0001) and both were associated with the DO intensity categories (AW: r(sp) = 0.47, AG: r(sp) = 0.47; p <.001). At the individual level, all devices demonstrated relatively low sensitivity but higher specificity. At the group level, AW-E and AG-P provided similar estimates of time spent in moderate-to-vigorous PA (MVPA, AW-E: 4.7 +/- 4.1, AG-P: 4.4 +/- 3.3), compared with DO (5.1 +/- 3.5). Conclusion: The AW-E and AG-P estimated times spent in MVPA were similar to DO, but the weak agreement statistics indicate that neither device cut-point equations provided accurate estimates at the individual level.